The impact of pharmaceutical innovation on cancer mortality in Mexico, 2003–2013

Abstract

I assess the impact that pharmaceutical innovation had on cancer mortality in Mexico during the period 2003–2013, by investigating whether there were larger declines in the age-standardized mortality rate of cancer sites (breast, lung, colon, etc.) that were subject to more pharmaceutical innovation, controlling for changes in the age-standardized cancer incidence rate. The estimates indicate that new drugs launched during 1991–2001 reduced the age-standardized cancer mortality rate by 16%, i.e., at an average annual rate of about 1.6%. I estimate that 105,661 life-years before age 70 were gained in 2013 due to cancer drugs launched during 1991–2001, and that the cost per life-year gained was in the neighborhood of $2146. By the standards of the World Health Organization, new cancer drugs have been very cost-effective in Mexico. The contribution of cancer drug innovation to Mexican longevity growth has been valuable, but, perhaps, it could have been even larger. Only half as many new cancer drugs were launched in Mexico during 2010–2014 as were launched in the US. In addition, when new drugs are launched in Mexico, their diffusion tends to be quite slow.

Introduction

Cancer mortality has declined in Mexico during the last 2 decades. As shown in Fig. 1, the age-standardized cancer mortality rate¹ of males declined by 13%, and that of females declined by 11%, between 1995 and 2014.² The mortality rate, or (unconditional) probability of death from cancer, depends to an important extent on two variables: the probability of getting (being diagnosed with) cancer, and the probability of dying from cancer, conditional on having been diagnosed with cancer: prob(death) ≈ prob(diagnosis) * prob(death|diagnosis).³ Therefore, the decline in the mortality rate could be due to either a decline in cancer incidence⁴, a decline in the probability of dying from cancer, conditional on having been diagnosed with cancer (e.g., due to improved treatment), or both.

Fig. 1

(Source: WHO Cancer Mortality database, http://www-dep.iarc.fr/WHOdb/WHOdb.htm)

Age-standardized cancer mortality rates, by sex, Mexico, 1995–2014

The Mexican cancer incidence rate declined by 13% (from 147.3 to 128.4) between 2002 and 2008, although it increased 2% (from 128.4 to 131.5) between 2008 and 2012.⁵ Since the cancer incidence rate declined at least as rapidly as the cancer mortality rate in recent years, the decline in cancer mortality could be entirely due to declining cancer incidence. However, the measurement of cancer incidence is subject to significant potential errors. For example, a decline in cancer surveillance or screening could lead to a decline in measured cancer incidence, even when true incidence is not declining.

The previous studies (Lichtenberg 2014a, 2015, 2016a, b) have shown that pharmaceutical innovation—the introduction and use of new cancer drugs—has significantly reduced cancer mortality in countries at a “very high” level of human development (as defined by the United Nations Development Programme⁶). In this study, I will assess the impact that pharmaceutical innovation and cancer incidence had on cancer mortality in Mexico, a country at a lower (but still “high”) level of human development. As in the previous studies, a difference-in-difference research design will be used: I will investigate whether the decline in mortality was greater for cancer sites (breast, lung, colon, etc.) subject to more pharmaceutical innovation and greater declines in incidence. As shown in Fig. 2, the rate of decline in the mortality rate varied considerably across cancer sites. The mortality rate declined by at least 34% for 3 cancer sites (cervix, stomach, and lung), but increased for 3 other cancer sites (colon, ovary, and breast).

Fig. 2

(Source: OECD Health Statistics 2016 database)

% change in age-adjusted mortality rate, by cancer site, 2000–2013

In Sect. 2, I will formulate an econometric model of cancer mortality. The data sources used to estimate these models are described in Sect. 3. Empirical results are presented in Sect. 4. Rough estimates of the number of life-years gained in 2013 from the reduction in cancer mortality attributable to pharmaceutical innovation, and of the average cost-effectiveness (cost per life-year gained) of new cancer drugs, are developed in Sect. 5. Sect. 6 provides a summary and conclusions.

Econometric model of cancer mortality

The basic model which I will use to assess the impact of pharmaceutical innovation and cancer incidence on age-standardized cancer mortality rates in Mexico is:

$$ \begin{aligned} {\text{MORT}}_{st} = \beta_{k} {\text{CUM}}\_{\text{NCE}}_{s,t - k} + \gamma \,{\text{INCIDENCE}}_{s,t - 1} + \alpha_{s} + \delta_{t} + \varepsilon_{st} , \\ \end{aligned} $$

(1)

where MORT _st  = the age-standardized mortality rate from cancer at site s in year t (t = 2003, 2013); CUM_NCE _s,t–k  = ∑ _d IND _ds LAUNCHED _d,t−k = the number of new chemical entities (drugs) to treat cancer at site s that had been launched in Mexico by the end of year t−k (k = 0, 3, 6 ,…,18); ${\text{IND}}_{ds} \begin{array}{ll} {{ = }\; 1 {\text{ if drug }}d{\text{ is used to treat }}\left( {\text{indicated for}} \right){\text{ cancer at site }}s} \\ { = 0{\text{ if drug }}d{\text{ is not used to treat }}\left( {\text{indicated for}} \right){\text{ cancer at site }}s} \\ \end{array} ;$ ${\text{LAUNCHED}}_{d,t - k} \begin{array}{*{20}c} {{ = }\, 1 {\text{ if drug }}d{\text{ had been launched in Mexico by the end of year }}t - k} \\ {{ = }\,0{\text{ if drug }}d{\text{ had not been launched in Mexico by the end of year }}t - k} \\ \end{array} ;$ INCIDENCE _s,t−1 = the age-standardized incidence rate of cancer at site s in year t − 1; α _s  = a fixed effect for cancer at site s; δ _t  = a fixed effect for year t.

Inclusion of year and cancer-site fixed effects controls for the overall decline in cancer mortality and for stable between-cancer-site differences in mortality. Negative and significant estimates of β _k in Eq. (1) would signify that cancer sites for which there was more pharmaceutical innovation had larger declines in mortality, controlling for changes in incidence.

Due to data limitations, the number of new chemical entities is the only cancer-site-specific, time-varying, measure of medical innovation in Eq. (1). Both a patient-level US study and a longitudinal country-level study have shown that controlling for numerous other potential determinants of mortality does not reduce, and may even increase, the estimated effect of pharmaceutical innovation. The study based on patient-level data (Lichtenberg 2013) found that controlling for race, education, family income, insurance coverage, Census region, BMI, smoking, the mean year the person started taking his or her medications, and over 100 medical conditions had virtually no effect on the estimate of the effect of pharmaceutical innovation (the change in drug vintage) on life expectancy. The study based on longitudinal country-level data (Lichtenberg 2014b) found that controlling for ten other potential determinants of longevity change [real per capita income, the unemployment rate, mean years of schooling, the urbanization rate, real per capita health expenditure (public and private), the DPT immunization rate among children ages 12–23 months, HIV prevalence, and tuberculosis incidence] increased the coefficient on pharmaceutical innovation by about 32%.

Failure to control for non-pharmaceutical medical innovation (e.g., innovation in diagnostic imaging, surgical procedures, and medical devices) is also unlikely to bias estimates of the effect of pharmaceutical innovation on premature mortality, for two reasons. First, more than half of US funding for biomedical research came from pharmaceutical and biotechnology firms (Dorsey et al. 2010). Much of the rest came from the federal government (i.e., the NIH), and new drugs often build on upstream government research (Sampat and Lichtenberg 2011). The National Cancer Institute (2016a, b) says that it “has played an active role in the development of drugs for cancer treatment for 50 years… [and] that approximately one half of the chemotherapeutic drugs currently used by oncologists for cancer treatment were discovered and/or developed” at the National Cancer Institute. Second, the previous research based on US data (Lichtenberg 2014a, c) indicates that non-pharmaceutical medical innovation is not positively correlated across diseases with pharmaceutical innovation. However, while non-pharmaceutical medical innovation may not be correlated with pharmaceutical innovation across diseases in the US, this need not hold for Mexico.

The measure of pharmaceutical innovation in Eq. (1)—the number of chemical substances previously registered to treat cancer at site s—is not the theoretically ideal measure. Mortality is presumably more strongly related to the drugs actually used to treat cancer than it is to the drugs that could be used to treat cancer. A preferable measure is the mean vintage of drugs used to treat cancer at site s in year t, defined as VINTAGE _st  = ∑ _d Q _dst LAUNCH_YEAR _d /∑ _d Q _dst , where Q _dst  = the quantity of drug d used to treat cancer at site s in year t, and LAUNCH_YEAR _d  = the world launch year of drug d.⁷ Unfortunately, measurement of VINTAGE _st is infeasible: even though data on the total quantity of each drug in each year (Q _d.t = Σ _s Q _dst ) are available, many drugs are used to treat multiple diseases. There is no way to determine the quantity of drug d used to treat cancer at site s in year t.⁸ However, Lichtenberg (2014c) showed that in France, there is a highly significant positive correlation across drug classes between changes in the (quantity-weighted) vintage of drugs and changes in the number of chemical substances previously registered within the drug class.

In principle, it could be desirable to control for the length of time between the ‘world launch’ of cancer drugs and their ‘Mexico launch’.⁹ According to Solow’s vintage hypothesis, later, vintage goods (e.g., drugs whose world launch years were later) are likely to be of higher quality than earlier vintage goods. Holding constant the Mexican launch year of a drug, the shorter the lag from world launch year to Mexican launch year, the later the world launch year of the drug, and (according to the vintage hypothesis), the higher the drug’s quality. However, controlling for the length of time between the ‘world launch’ of cancer drugs and their ‘Mexico launch’ is problematic. I can compute the mean lag between the world launch year and the Mexican launch year for cancer sites and years in which at least one drug had been launched in Mexico. However, I cannot compute the mean lag for cancer sites and years in which no drugs had been launched in Mexico. As shown in Table 2, 8 cancer sites had 0 drug launches by 1995; 4 cancer sites had 0 drugs launches by 2004. Controlling for the length of time between the ‘world launch’ of cancer drugs and their ‘Mexico launch’ would require excluding those observations.

In Eq. (1), mortality from cancer at site s in year t depends on the number of new chemical entities (drugs) to treat cancer at site s that had been launched in Mexico by the end of year t − k, i.e., there is a lag of k years. Equation (1) will be estimated for different values of k: k = 0, 3, 6,…,18. A separate model is estimated for each value of k, rather than including multiple values (CUM_NCE _s,t , CUM_NCE _s,t−3, CUM_NCE _s,t−6,…) in a single model, because CUM_NCE is highly serially correlated (by construction), which would result in extremely high multicollinearity if multiple values were included. One would expect there to be a substantial lag, because new drugs diffuse gradually—they will not be used widely until years after registration. Data from the IMS Health MIDAS database can be used to provide evidence about the process of diffusion of new medicines. I used data from that source linked to data on Mexican drug launch dates (described below) to estimate the following model:

$$ { \ln }\left( {{\text{N}}\_{\text{RX}}_{dy} } \right) \, = \rho_{d} + \pi_{y} + \varepsilon_{dy} , $$

(2)

where N_RX _dy = the number of standard units of cancer drug d sold in Mexico per thousand population y years after it was launched (y = 0–3, 4–7, 8–11, 12–15, 16–19 years); ρ _d  = a fixed effect for drug d; π _y = a fixed effect for age y.

The expression $\exp \left( {\pi_{y} - \pi_{16 - 19} } \right)$ is a “relative utilization index”: it is the mean ratio of the annual number of standard units of a cancer drug sold per thousand population y years after it was first launched in Mexico to the annual number of standard units of the same drug sold per thousand population 16–19 years after it was first launched in Mexico.

Using annual data on the number of standard units of cancer drugs sold in Mexico during the period 1999–2010, I estimated Eq. (2). Estimates of the “relative utilization index” are shown in Fig. 3. These estimates indicate that utilization of a drug is strongly positively related to how long the drug has been on the market. On average, a drug is used 50 times as often per year 16–19 years post-launch as it is 0–3 year post-launch. Utilization appears to rise especially rapidly after year 11, when drug patents tend to expire and generics enter the market.

Fig. 3

Relative utilization of cancer drugs in Mexico, by number of years since launch (index: ratio of utilization y years since launch to utilization 16–19 years since launch)

The effect of a drug’s launch on mortality is likely to depend on both the quality and the quantity of the drug. Indeed, it is likely to depend on the interaction between quality and quantity: a quality improvement will have a greater impact on mortality if drug utilization (quantity) is high. Although newer drugs tend to be of higher quality than older drugs (see Lichtenberg 2014d), the relative quantity of very new drugs is quite low, so the impact on mortality of very new drugs is lower than the impact of older drugs.

In principle, mortality in year t should depend on a distributed lag of incidence, i.e., on INCIDENCE _s,t , INCIDENCE _s,t−1, INCIDENCE _s,t−2, INCIDENCE _s,t−3… Unfortunately, data on incidence by cancer site are available for only 2 years (2002 and 2012); this is why INCIDENCE _s,t−2 is the only incidence variable included in Eq. (1). The limited availability of incidence data also means that we can only use mortality data for 2 years (2004 and 2014). Writing the model for each of these years:

$$ \begin{aligned} {\text{MORT}}_{s,2003} = \beta_{k} {\text{CUM}}\_{\text{NCE}}_{s, 200 3- k} + \gamma {\text{INCIDENCE}}_{s, 200 2} + \alpha_{s} + \delta_{ 200 3} + \varepsilon_{s,2003} , \\ \end{aligned} $$

(3)

$$ \begin{aligned} {\text{MORT}}_{s, 20 1 3} = \beta_{k} {\text{CUM}}\_{\text{NCE}}_{s,2013 - k} \ + \gamma \,{\text{INCIDENCE}}_{s,2012} + \alpha_{s} + \delta_{ 20 1 3} + \varepsilon_{s,2013} . \\ \end{aligned} $$

(4)

Subtracting (3) from (4),

$$ \begin{aligned} \left( {{\text{MORT}}_{s, 20 1 3} - {\text{ MORT}}_{s, 200 3} } \right) \, & = \beta_{k} \left( {{\text{CUM}}\_{\text{NCE}}_{s, 20 1 3- k} - {\text{ CUM}}\_{\text{NCE}}_{s, 200 3- k} } \right) \\ & \quad + \gamma \left( {{\text{INCIDENCE}}_{s, 20 1 2} - {\text{ INCIDENCE}}_{s,2002} } \right) \\ \, & \quad + \, (\delta_{ 20 1 3} - \delta_{ 200 3} ) \, + \, (\varepsilon_{s,2013} - \varepsilon_{s,2003} ). \\ \end{aligned} $$

(5)

Equation (5) may be rewritten as follows:

$$ \begin{aligned} \Delta {\text{MORT}}_{s} & = \beta_{k} \Delta {\text{CUM}}\_{\text{NCE}}\_k_{s} \\ & \quad + \gamma \,\Delta {\text{INCIDENCE}}\_ 1_{s} + \delta^{\prime} \, + \varepsilon_{s}^{\prime } , \\ \end{aligned} $$

(6)

where ∆MORT _s  = MORT _s,2013 − MORT _s,2003 = the 2003–2013 change in the age-standardized mortality rate from cancer at site s; ∆CUM_NCE_k _s  = CUM_NCE _s,2013k  − CUM_NCE _s,2003−k  = the number of drugs for cancer at site s launched between year 2003 − k and 2013 − k; ∆INCIDENCE_1 _s  = INCIDENCE _s,2012 − INCIDENCE _s,2002 = the 2002–2012 change in the age-standardized incidence rate of cancer at site s; δ′ = δ ₂₀₁₃ − δ ₂₀₀₃.

Equation (6) indicates that the 2003–2013 change in the age-standardized mortality rate depends on two variables: the number of drugs launched between year 2003 − k and 2013 − k, and the 2002–2012 change in the age-standardized incidence rate.

For estimates of β _k from Eqs. (1) and (6) to be consistent estimates of the effect of drug launches on mortality, the “parallel trends” assumption needs to be satisfied. A simple way to test the validity of this assumption is to estimate a version of Eq. (6) that includes a control for the trend in mortality in the ‘pre-period’, e.g., the period 1993–2003. Therefore, I will estimate the following model:

$$ \begin{aligned} \Delta {\text{MORT}}_{s} & = \beta_{k} \Delta {\text{CUM}}\_{\text{NCE}}\_k_{s} + \gamma \,\Delta {\text{INCIDENCE}}\_ 1_{s} \\ & \quad + \pi \, \Delta {\text{MORT}}\_{\text{PRE}}_{s} + \delta^{\prime} \, + \varepsilon_{s}^{\prime } , \\ \end{aligned} $$

(7)

where ∆MORT_PRE _s  = MORT _s,2003 − MORT _s,1993 = the 1993–2003 change in the age-standardized mortality rate from cancer at site s.

Data sources

Age-standardized cancer mortality rate data were obtained from the WHO Cancer Mortality database (World Health Organization (2016a)).

Age-standardized cancer incidence rate data were obtained from GLOBOCAN International Agency for Research on Cancer (2016). Mortality and incidence data are reported separately by sex. For cancers affecting both sexes, we computed the simple mean of the sex-specific rates. For cancers affecting only one sex (breast, cervical, ovarian, and prostate), we computed 50% of the single-sex rate.

Data on drugs approved for different types of cancer were obtained from the US National Cancer Institute.

Data on Mexican launch dates of drugs were obtained from the IMS Health New Product Focus database. This database contains data on drug launches (in Mexico and many other countries) from 1982 to the present. The Mexican launch year (as indicated in the IMS New Product Focus database) is usually the year in which the drug was first sold in Mexico (as indicated in the IMS MIDAS database). In a few cases, the drug was first sold in Mexico in the year after the Mexican launch year. I define CUM_NCE _st as the number of post-1981 new chemical entities (i.e., NCEs first launched anywhere in the world after 1981) used to treat cancer at site s that had been launched in Mexico by the end of year t. Since the New Product Focus data are left-censored (no pre-1982 data), this measure is subject to error, because CUM_NCE _st will not (but should) include pre-1982 NCEs that were first launched in Mexico after 1981. If this measurement error is random, it is likely to bias estimates of β _k towards zero.

Annual data on the number of standard units of cancer drugs sold in Mexico during the period 1999–2010 were obtained from the IMS Health MIDAS database.

Data on age-standardized mortality and incidence rates, by cancer site and year, are shown in Table 1.

Table 1 Age-standardized mortality and incidence rates, by cancer site, 1993–2013

Data on the number of post-1981 drugs ever launched in Mexico, by cancer site and year, are shown in Table 2.

Table 2 Number of post-1981 drugs ever launched in Mexico, by cancer site, 1986–2013

Mexican launch dates of drugs used to treat different types of cancer are shown in Appendix Table 5.

Empirical results

Estimates of the pharmaceutical innovation (β _k ) parameters of the cancer mortality rate model (Eq. (1) for different values of k are shown in Table 3. Each estimate is from a separate model. For simplicity, estimates of the incidence coefficient (γ) are not shown here. Estimates of the incidence coefficient were positive and significant (and virtually identical) in all models, indicating that mortality declined more for cancer sites that had larger declines in incidence.

Table 3 Estimates of β _k parameters of Eq. (1), MORT _st  = β _k CUM_NCE _s,t−k + γ INCIDENCE _s,t−1 + α _s + δ _t + ε _st for different values of k

The first seven rows of Table 3 show estimates based on mortality and incidence data for both sexes combined. Row 1 shows the estimate of β ₀, i.e., of the pharmaceutical innovation coefficient when there is no lag between the number of drugs ever launched and the mortality rate. The estimate of β ₀ is negative but only marginally statistically significant (p = 0.07). This is not surprising, since, as shown in Fig. 3, utilization of recently launched drugs tends to be quite low.

Row 2 shows the estimate of β ₃, i.e., of the pharmaceutical innovation coefficient when there is a 3-year lag between the number of drugs ever launched and the mortality rate. The estimate of β ₃ is negative and statistically significant (p = 0.03). Rows 3–7 of Table 3 show estimates of β _k for k = 6, 9, 12, 15, and 18, respectively. All five β _k estimates are negative and statistically significant (p ≤ 0.031). These estimates signify that the mortality rate is inversely related to the number of drugs that had ever been launched 3–18 years earlier, controlling for the incidence rate.¹⁰ The estimates of β _k based on data for both sexes combined are plotted in Panel A of Fig. 4.

Fig. 4

Estimates of β _k parameters of Eq. (1), MORT _st  = β _k CUM_NCE _{s, t−k} + γ INCIDENCE _{s, t−1} + α _s + δ _t + ε _st for different values of k. a Both sexes, b females, c males

To calculate the change in the mean mortality rate attributable to pharmaceutical innovation, we can multiply the estimate of β _k (column 1 of Table 3) by the mean value of ∆CUM_NCE_k (column 5); the result is shown in column 6. These calculations imply that new drugs launched during the period 1991–2001 had the largest (most negative) effect on the 2003–2011 change in mortality. Henceforth, I will focus on the estimates of the model when k = 12.

The two explanatory variables (∆CUM_NCE_12 and ∆INCIDENCE_1) in the mortality change (∆MORT) model jointly explain 77% of the variance across cancer sites in the 2003–2013 change in the mortality rate. The mean mortality rate declined by 0.47 between 2003 and 2013. The estimates imply that new drugs launched during 1991–2001 reduced the mean mortality rate by 0.47. The mean mortality rate in 2003 was 2.99, so this indicates that new drugs launched during 1998–2008 reduced the mean mortality rate by 16% (= 0.47/2.99), i.e., at an average annual rate of about 1.6%. The decline in incidence also reduced the mean mortality rate, but by only 60% as much as pharmaceutical innovation (0.28). The estimates also indicate that if no new drugs had been launched during 1991–2001 and the incidence rate had not declined between 2002 and 2012, the mortality rate would have increased by 0.28 between 2003 and 2013, although the estimated increase (the intercept) is not statistically significant.

Rows 8–14 of Table 3 show estimates based on mortality and incidence data for females only. The estimates are not statistically significant when k ≤ 6, but 3 of the 4 estimates for k ≥ 9 are negative and statistically significant. These estimates are plotted in Panel B of Fig. 4.

Rows 15–21 of Table 3 show estimates based on mortality and incidence data for males only. The estimates are not statistically significant when k ≤ 9, but all three estimates for k ≥ 12 are negative and statistically significant. These estimates are plotted in Panel C of Fig. 4. The difference between the effects of pharmaceutical innovation on female and male cancer mortality rates is not statistically significant. For example, the p value on the difference between the female and male estimates of β ₁₂ in Table 3 (0.0972 = − 0.2218 to (− 0.3191)) is 0.51 (Chi-square = 0.44).

Launching of new drugs in Mexico may not be strictly exogenous with respect to Mexican cancer mortality. To address the potential endogeneity of drug launches in Mexico, I estimated a version of Eq. (1) via instrumental variables (IV). The instrument that I used for CUM_NCE _s,t−k (the number of drugs to treat cancer at site s that had been launched in Mexico by the end of year t−k) is CUM_NCE_FOREIGN _s,t−k (the number of drugs to treat cancer at site s that had been launched outside of Mexico by the end of year t−k), defined as follows: CUM_NCE_FOREIGN_s,t−k  = ∑ _d IND _ds LAUNCHED_FOREIGN _d,t−k  = the number of new chemical entities (drugs) to treat cancer at site s that had been launched outside of Mexico by the end of year t−k (k = 0, 3, 6,…,18): ${\text{LAUNCHED}}\_{\text{FOREIGN}}_{d,t - k} \begin{array}{*{20}c} { = {\text{ 1 if drug }}d{\text{ had been launched}}\;outside \, of\;{\text{Mexico by the end of year }}t - k} \\ { = \, 0{\text{ if drug }}d{\text{ had not been launched}}\;outside \, of\;{\text{Mexico by the end of year }}t - k} \\ \end{array} .$

The first stage of the two-stage IV estimation procedure is to estimate the following equation:

$$ {\text{CUM}}\_{\text{NCE}}_{{s,t\text{ - }k}} = \, \varOmega {\text{ CUM}}\_{\text{NCE}}\_{\text{FOREIGN}}_{s,t - k} + \alpha_{s} + \delta_{t} + \varepsilon_{st} . $$

(8)

CUM_NCE_FOREIGN _s,t−k is a good instrument for CUM_NCE _s,t−k : when k = 12, the estimate of Ω is 0.8368 (Z = 6.24; p < 0.0001). The second stage is to estimate Eq. (1), replacing the actual value of CUM_NCE _s,t−k by its predicted value from Eq. (8). The IV estimate of β ₁₂ for both sexes (IV estimate = 0.2329; Z = 4.06; p < 0.0001) is very similar to the OLS estimate shown in row 5 of Table 3.

As discussed earlier, the “parallel trends” assumption needs to be satisfied for the estimates in Table 3 and Fig. 4 to be consistent estimates of the effect of drug launches on mortality. To provide evidence about the validity of this assumption, in Table 4, I present estimates of Eq. (7), which includes a control for the trend in mortality in the ‘pre-period’, e.g., the period 1993–2003. These estimates are based on a smaller set of cancer sites, because, as shown in Table 1, data on age-standardized mortality rates are less complete in 1993 than they are in 2003 and 2013. Rows 22–28 of Table 4 show estimates of Eq. (7) based on data for both sexes combined for this smaller set of cancer sites when the trend in mortality in the ‘pre-period’ (1993–2003) is excluded from the equation. Rows 29–35 show estimates of Eq. (7) based on the same sample when the trend in mortality in the ‘pre-period’ is included in the equation. The coefficient (π) on the pre-period mortality trend is significant (and positive) only when k ≥ 15. Controlling for the pre-period mortality trend does not substantially change the estimates of β _k . The 2003–2013 change in the age-standardized mortality rate is inversely related to the change in the number of drugs ever launched 12–18 years earlier, controlling for the 2002–2012 change in incidence and the ‘pre-period’ (1993–2003) change in the mortality rate.

Table 4 Estimates of β _k parameters of Eq. (7), ∆MORT _s = β _k ∆CUM_NCE_k _s + γ ∆INCIDENCE_1 _s + π ∆MORT_PRE _s + δ′ + ε _s ′ excluding and including ∆MORT_PRE _s

The relationship across cancer sites between the number of new drugs launched during 1985–1995 and the 2003–2013 change in the mortality rate, controlling for the “pre-period” (1993–2003) change in the mortality rate and the 2002–2012 change in the incidence rate, is shown in Fig. 5; this chart corresponds to the estimate of β ₁₈ shown in row 35 of Table 4. The figure indicates that the three cancer sites with the largest number of new drugs launched during 1985–1995 had the largest 2003–2013 reductions in the mortality rate, controlling for the 2002–2012 change in the incidence rate and the trend in mortality in the pre-period. Excluding these three cancer sites from the sample does not have much effect on the estimate of β ₁₈; when they are excluded, the estimate of β ₁₈ is − 0.357 (t = 8.15; p = 0.0039).

Fig. 5

Relationship across cancer sites between the number of new drugs launched during 1985–1995 and the 2003–2013 change in the mortality rate, controlling for the 1993–2003 change in the mortality rate and the 2002–2012 change in the incidence rate

Discussion

The estimates indicate that the launch of new drugs subsequently reduced cancer mortality. New drugs launched during 1991–2001 are estimated to have reduced the age-standardized cancer mortality rate by 16%, i.e., at an average annual rate of about 1.6%.

Now, I will develop a rough estimate of the number of life-years gained in 2013 from the reduction in cancer mortality attributable to pharmaceutical innovation, and of the average cost-effectiveness (cost per life-year gained) of new cancer drugs.

To do this, I will estimate the decline in the premature mortality rate attributable to pharmaceutical innovation. The premature mortality rate is the number of potential years of life lost (PYLL) per 100,000 population (OECD 2017a, b). PYLL is a summary measure of premature mortality that provides an explicit way of weighting deaths occurring at younger ages. The calculation of PYLL involves summing up deaths occurring at each age and multiplying this with the number of remaining years to live up to a selected age limit.¹¹ The limit of 70 years was chosen for the calculations in OECD Health Statistics. To assure cross-country and trend comparison, the PYLL are standardized, for each country and each year. The total OECD population in 2010 is taken as the reference population for age standardization.

As shown in Table 1, between 2003 and 2013, the age-standardized mortality rate for all cancers combined declined by 15%, from 73.0 to 62.3. The estimates in Table 4 imply that virtually, this entire decline was due to the previous launches of new cancer drugs. During the same period, according to the OECD, the premature cancer mortality rate (the number of PYLL before age 70/100,000 population below age 70) declined by 11.4%, from 782.4 to 693.4. It seems reasonable to assume that this entire decline was also due to the previous launches of new cancer drugs. Therefore, in the absence of the previous new drug launches, premature cancer mortality would have been 12.8% (= (1/(1 − 0.114)) − 1) higher in 2013 than it actually was. Actual PYLL before age 70 due to cancer in 2013 was 823,209 (= 693.4 * (118,720,632/100,000)). I estimate that in the absence of previous new drug launches, premature cancer mortality would have been 928,870 (= 112.8% * 823,209).

This calculation implies that 105,661 life-years before age 70 were gained in 2013 due to new cancer drugs. This is a rough estimate of the longevity benefit in 2013 to people under 70 of cancer drugs launched during the period 1991–2001. To calculate the average cost-effectiveness of these drugs, I would like to measure expenditure in 2013 by (or on behalf of) people under 70 on cancer drugs launched during the period 1991–2001. Unfortunately, these data are not available. However, I do have unpublished data from the IMS Health MIDAS database on expenditure (by or on behalf of all patients) by drug in 2010. Expenditure in 2010 on cancer drugs launched during 1991–2001 was $315 million.¹² 72% of people diagnosed with cancer in 2012 were below age 70 (source: GLOBOCAN). Expenditure in 2010 by or on behalf of people below age 70 on cancer drugs launched during 1994–2004 may, therefore, have been $227 million (= 72% * $315 million).

These calculations imply that the cost per life-year gained by people below age 70 from new cancer drugs was in the neighborhood of $2146 (= $227 million/105,661 life-years). This figure may be somewhat underestimated, since it is based on 2010 expenditure data.¹³ On the other hand, Lichtenberg (2014a) showed that in the US, about 25% of the cost of new drugs (for all diseases) tends to be offset by reduced expenditure on old drugs, so the cost per life-year gained may have been below $2000.

The World Health Organization considers interventions whose cost per quality-adjusted life-year (QALY) gained is less than per capita GDP to be “very cost-effective” (Bertram et al. 2016); Mexico’s per capita GDP in 2011 was $10,307.¹⁴ The estimated cost per life-year gained from the previous pharmaceutical innovation is also well below the vast majority of estimates from the value-of-life literature of the value of a life-year (see Hirth et al. 2000).

Summary and conclusions

I assessed the impact that pharmaceutical innovation had on cancer mortality in Mexico during the period 2003–2013, by investigating whether there were larger declines in mortality for cancer sites (breast, lung, colon, etc.) that were subject to more pharmaceutical innovation, controlling for changes in cancer incidence. New drugs launched during 1991–2001 are estimated to have reduced the age-standardized cancer mortality rate by 16%, i.e., at an average annual rate of about 1.6%. I estimated that 105,661 life-years before age 70 were gained in 2013 due to cancer drugs launched during 1997–2007, and that the cost per life-year gained was in the neighborhood of $2146. By the standards of the World Health Organization, new cancer drugs have been very cost-effective in Mexico.

The contribution of cancer drug innovation to Mexican longevity growth has been valuable, but, perhaps, it could have been even larger. According to the IMS Institute for Healthcare Informatics (2016), during the period 2010–2014, 49 new cancer medicines were launched worldwide. As shown in Fig. 6, about twice as many of these have been launched in the US as have been launched in Mexico.¹⁵ In addition, as shown in Fig. 3, when new drugs are launched in Mexico, their diffusion is quite slow.

Fig. 6

Number of 2010–2014 cancer medicines that have been launched in various regions

Due to unavailability of data, this study is subject to several limitations. The measure of pharmaceutical innovation—the number of chemical substances previously launched to treat cancer—is not the theoretically ideal measure. The number of chemical substances previously launched was the only cancer-site-specific, time-varying, measure of medical innovation. The previous research based on US data indicates that non-pharmaceutical medical innovation is not positively correlated across diseases with pharmaceutical innovation, but this may not apply to Mexico. Future research may be able to overcome these and other limitations.

Appendix

See Table 5.

Table 5 Mexican launch dates of drugs used to treat different types of cancer